Christie Mealo is Chief AI Officer at Medical Knowledge Group, where she leads AI strategy and helps scale AI across the organisation and its client solutions. With more than a decade of experience building AI driven products in healthcare and life sciences, she brings deep expertise in product innovation, enterprise transformation, and responsible AI leadership.

A frequent speaker on AI strategy and the author of The Probabilistic Product, in the interview below she shares how AI can help life sciences organisations improve clinical adoption and turn innovation into real world impact.

Interview:

1. How are current drug development pipelines different today than they were in the not too distant past, and how has that changed what life sciences companies need to do to support clinical and commercial adoption?

CM: Drug pipelines today look very different than they did even a decade ago. We still have medicines designed for large patient populations, but portfolios now include far more targeted therapies across oncology, rare disease, and other complex conditions. Many involve novel mechanisms of action, biomarker testing, specialised administration, or a care pathway that crosses several clinicians and treatment settings.

That matters because the work after approval has changed too. A clinician may need to understand the underlying science, identify a very specific patient, order the right test, coordinate a referral, and work through access or reimbursement questions before treatment begins.

Broad consumer advertising can still create awareness, but it cannot carry that burden by itself. A Super Bowl ad cannot explain a new mechanism of action or help a physician know when to refer a patient to a specialist. Clinical and commercial success increasingly depends on reaching the right clinicians with information they can actually use, at the point when it can change care.

2. What does the diffusion of innovation mean in healthcare, and why is it becoming increasingly important?

CM: I use “diffusion of innovation” fairly literally. It is the process by which a new treatment moves from an approved scientific innovation into appropriate, sustained use in clinical practice. Approval is a critical milestone, but it does not mean that every clinician immediately understands where the treatment fits, which patients may benefit, or what has to change in the existing care pathway.

Adoption usually moves through people and networks. Investigators and recognised experts understand the evidence first; early adopters begin applying it; trusted peers, referral relationships, guidelines, and institutional protocols help it spread more broadly.

The more complex the mechanism of action or treatment pathway, the more intentional that process needs to be. We need to understand who can interpret the science, who can identify the patient, who influences practice, and where confusion or friction is slowing adoption. That is the work behind diffusion, and it is becoming central to how life sciences companies support both clinical and commercial success.

3. How can AI help life sciences companies identify the right clinicians, communicate complex evidence, and support changes in clinical practice?

CM: AI gives us a better way to understand how clinical influence and adoption move, provided the information is appropriately sourced, permissioned, and used. Signals from publications, clinical trials, congress activity, referral networks, and approved engagement data can help teams see who is shaping a field, where knowledge gaps exist, and which clinicians may need a different kind of scientific conversation.

AI can also help qualified teams synthesise new evidence, identify recurring questions, and prepare approved scientific material for different audiences. A recognised expert who already understands the mechanism of action does not need the same information as a physician encountering the treatment for the first time.

I am less interested in using AI simply to generate more content. The useful question is whether it helps experts communicate with greater precision and respond faster to what clinicians actually need. Medical, regulatory, and clinical professionals still own the evidence, the judgement, and the final communication. AI helps them see patterns and apply that expertise more effectively.

4. At Medical Knowledge Group, you began by analysing individual tasks and workflows through the AI Advocates program. What did that reveal about how organisations should approach AI transformation?

CM: Our AI Advocates work reinforced something I have believed for a long time: enterprise change requires both top down commitment and bottom up design. Leadership has to establish the direction, provide resources, set guardrails, and make it clear that the work matters. The people closest to the work, however, are the ones who can show you where time is being lost and where quality, consistency, or service can improve.

During one initial two month assessment, we identified more than 50 tasks across eight departments. That specificity mattered because it moved the conversation beyond generic ideas about AI and into the actual work people were doing every day.

I have been calling this grassroots enterprise engineering. The transformation team becomes the connective tissue between leadership and employees, turning local insight into repeatable tools and better workflows. The goal is not simply to remove cost. It is to reinvest capacity in stronger thinking, better output, faster response, and a higher level of service for our clients.

5. How has MKG progressed from general purpose AI tools and point solutions toward proprietary products and connected workflows?

CM: We did not begin by declaring that we needed one large AI platform. We started with the work. The AI Advocates program helped us map tasks, friction, quality risks, and opportunities across the company. From there, we built repeatable Claude Projects, agents, prompts, templates, and other focused tools around patterns that were already proving useful.

The next step was to create products for specific workflows we understand well. Vector supports proof review, routing, and approvals, while Annotation Activation supports our medical teams.

We are now connecting those capabilities through MKG1 and the MKG Mesh. MKG1 is the unified experience we are building for accessing knowledge, expertise, agents, and workflow tools. The MKG Mesh is the permissioned technology underneath it, designed to connect knowledge graphs, agents, workflows, and enterprise systems. This architecture lets us build a strong MKG experience while making our best knowledge easier to use. The ultimate goal is better work and better service for our clients.

6. Your book, The Probabilistic Product, focuses on building differentiated and defensible AI native businesses. How has that framework shaped the choices you are making at MKG?

CM: The central argument in The Probabilistic Product is that access to a powerful model is not a defensible advantage. Most companies can access the same frontier models, and those models will continue to improve and become more widely available. The harder question is what you build around them.

I frame that through three decisions: define, differentiate, and defend. You have to define the value and boundaries of the product, differentiate it through knowledge, workflow, experience, distribution, or trust, and then build the evaluation and learning systems that allow the product to improve through use.

That framework has shaped how I think about our work at MKG. Vector, MKG1, and the MKG Mesh are grounded in life sciences work that we understand deeply. Rather than beginning with a general purpose chatbot, we are connecting our expertise, permissioned knowledge, and operating processes around specific work. The model is important, but the enduring value sits in the system, experience, and institutional knowledge built around it.

7. What becomes possible when an organisation connects its best work, employee expertise, and operational knowledge to a permissioned AI and agent layer?

CM: Like most organisations, MKG already has an enormous amount of valuable knowledge. The problem is that it lives across project files, SharePoint, email, research, prior work, and the experience of individual employees. People can spend far too much time looking for something the company already knows or trying to find the person who knows it.

MKG1 is the experience we are building to make that knowledge and capability useful in the flow of work. The MKG Mesh sits underneath it, connecting permissioned knowledge graphs, agents, tools, workflows, and enterprise systems. With the appropriate controls, the same approved capability can eventually be available across supported work surfaces.

Client separation, role based access, source provenance, and retention requirements need to be designed into that architecture from the beginning. We are not putting every piece of information into one communal pool. We are creating a controlled way for people to access approved knowledge in the right context, respond faster, and produce stronger work for clients.

8. As AI becomes more agentic, what controls and human accountability are required in a regulated industry such as healthcare?

CM: As systems become more agentic, governance has to extend beyond what an AI system can say to what it is allowed to do. Every agent needs a clear purpose, a responsible owner, bounded authority, and an escalation path when the situation falls outside that authority.

I think the level of human involvement should follow the consequence of the action. A low risk, reversible administrative task may support more autonomy. Scientific claims, medical interpretation, regulated content, and external communications require qualified review and established approval processes. Reading information, making a recommendation, routing work, and executing an action are also different levels of permission and should be treated that way.

Organisations need traceability across sources, actions, changes, and approvals, along with ongoing evaluation and a practical recovery process when something goes wrong. Agentic systems can behave in ways that are not always predictable, which makes monitoring and escalation essential. Execution may be delegated to an agent. Accountability for the outcome still belongs to the organisation.

Conclusion

Christie’s perspective highlights a critical shift for healthcare and life sciences leaders. The challenge is no longer simply generating innovation. It is ensuring that innovation reaches practice in a way that is precise, scalable, and responsible. From scientific engagement to workflow transformation and governance, AI has the potential to help organisations close that gap and deliver stronger outcomes across the enterprise.

To hear Christie expand on these ideas, join her session at The AI Summit New York: AI ROI Reality Check: Turning AI Investment into Business Impact, Wednesday, December 9, 5:05 PM, Beyond AI at Scale Stage.

Join Christie and an exciting panel of speakers for a practical discussion on how to turn AI investment into real business impact.